Learning Transferable Adversarial Robust Representations via Multi-view Consistency

TL;DR

This paper introduces a meta-adversarial multi-view learning framework with dual encoders that enhances the transferability and robustness of representations across unseen domains and tasks, significantly improving adversarial robustness in few-shot learning.

Contribution

It proposes a novel multi-view representation learning method with dual encoders and label-free adversarial attacks to improve generalizable adversarial robustness in meta-learning.

Findings

Over 10% robust accuracy improvements on unseen domain few-shot tasks.

Effective in learning transferable robust representations.

Outperforms previous adversarial meta-learning baselines.

Abstract

Despite the success on few-shot learning problems, most meta-learned models only focus on achieving good performance on clean examples and thus easily break down when given adversarially perturbed samples. While some recent works have shown that a combination of adversarial learning and meta-learning could enhance the robustness of a meta-learner against adversarial attacks, they fail to achieve generalizable adversarial robustness to unseen domains and tasks, which is the ultimate goal of meta-learning. To address this challenge, we propose a novel meta-adversarial multi-view representation learning framework with dual encoders. Specifically, we introduce the discrepancy across the two differently augmented samples of the same data instance by first updating the encoder parameters with them and further imposing a novel label-free adversarial attack to maximize their discrepancy. Then, we maximize the consistency across the views to learn transferable robust representations across domains and tasks. Through experimental validation on multiple benchmarks, we demonstrate the effectiveness of our framework on few-shot learning tasks from unseen domains, achieving over 10\% robust accuracy improvements against previous adversarial meta-learning baselines.